Agents on the inside, how AI copilots are reshaping the enterprise

Welcome to today’s edition of the Midas Report, where “business as usual” is quietly becoming “business with AI copilots, agents, and embedded intelligence everywhere.”

From ERP titan SAP going full agentic with Joule, to Workday deploying role based copilots across departments, the enterprise AI stack is shaping up fast. Meanwhile, McKinsey gives us a rare, sober look at where the returns (and risks) from gen AI are actually materializing, and Contextual AI is betting big on Retrieval Augmented Generation (RAG) to keep things grounded.

Let’s dive in.

SAP launches Joule Agents to automate enterprise muscle memory

SAP leveled up its enterprise AI ambitions this week, introducing Joule Agents, autonomous AI agents embedded throughout the SAP ecosystem to automate workflows, make decisions, and deliver some long overdue connective tissue across departments.

The announcement comes with depth, these aren’t generic assistants. SAP is positioning Joule Agents as business process savvy copilots, deeply integrated into SAP systems like S/4HANA, SuccessFactors, Concur, and more. Think of them less like a chatbot, more like a factory floor manager who also happens to live inside your cloud ERP.

SAP says the agents can slash workflow times by 90% in some cases (including transactions and navigation), reduce manufacturing defect inspections by 25%, improve fraud detection, and cut app development costs by nearly a third. Internal tools like Agent Builder and SAP Build Code are being marketed to developers to customize and scale these agents across organizations, a clear signal that SAP wants to be the platform, not just the provider.

For investors and operators, the key takeaway is scale. With 230 AI powered scenarios already shipping and a roadmap to 400 by end of 2025, SAP is making its AI push not just feature rich, but operationally central. If SAP’s model sticks, expect AI agents to become as ubiquitous in business process execution as APIs were in the last SaaS wave.

Workday enters the fray with role specific AI copilots

Not to be outdone, Workday is also charting an aggressive AI course, this time through the lens of job function. The company is shipping AI copilots that are finely tuned for HR, finance, legal, and IT teams, nudging the enterprise toward role aware intelligence rather than generalized bots.

This move matters not just for the functionality, it’s squarely targeting measurable ROI. Workday is betting that copilots that think like a recruiter or a financial controller are more likely to generate buy in (and real results) than generic assistants requiring constant human babysitting.

This tracks with broader trends. A McKinsey AI survey cited by Workday found nearly 80% of companies say they’re now using AI in at least one business function, up from 55% just a year ago. But maturity remains scarce, only 1% of organizations classify their AI deployments as “mature.” That gap between adoption and effectiveness is Workday’s opportunity.

Enterprises should watch closely as these role specific copilots roll out. If they streamline authority heavy, document intensive areas like legal or HR, Workday will push further into territory where Salesforce, SAP, and others have already planted AI flags. For teams planning their AI roadmaps, “function first” may soon be the preferred implementation strategy.

McKinsey offers an actual scoreboard for enterprise AI ROI

Amid endless talk of transformation and hype, McKinsey’s latest State of AI report snaps us all back to reality with a welcome question, where is AI delivering real business value?

According to the global survey of nearly 1,500 executives, returns are starting to materialize, but mostly when deployments are tightly paired with structural change. The most profitable companies aren’t just plugging in a model; they’re redesigning workflows, adopting KPI discipline, training roles in AI usage, and letting machines handle more narrow repeatable tasks.

Some noteworthy stats, 71% of companies now use gen AI, up from 33% in 2023. About 17% of respondents say gen AI contributed at least 5% to EBIT over the past year. And perhaps most telling, CEO oversight of AI initiatives has a strong correlation with ROI. In other words, senior commitment still makes a big difference.

However, growing pains remain. Nearly half of organizations reported at least one negative consequence from gen AI use (yes, hallucinations still haunt), and only 21% redesigned workflows, a crucial step if you want more than novelty demos. Organizations that skip foundational changes are finding that shiny AI tools don’t solve legacy process debt.

For decision makers, McKinsey’s bottom line is clear, AI pays when paired with operations, not just IT initiatives. If your gen AI deployment doesn't end with rethinking at least one workflow, you may be leaving ROI on the table.

Contextual AI goes all in on grounded intelligence

While the big platforms race to embed AI everywhere, Contextual AI is carving out a niche that may be just as critical, delivering Retrieval Augmented Generation (RAG) solutions tailored for enterprises trying to stay factual, and defensible.

The company specializes in building large scale knowledge apps anchored in trusted, internal documentation. Their pitch, trustworthy gen AI without the hallucination headaches. For regulated sectors like finance, healthcare, and government, that promise is a big deal, especially as risk and compliance teams start weighing in more heavily on AI deployments.

Contextual’s systems are already serving Fortune 500 clients, and their focus is less on hallucination prone text generation and more on grounded answers from vetted corpora. As more enterprise buyers move from experimentation toward “production grade AI,” expect RAG to become a must have architecture ingredient, not a science project.

We’re watching this space because grounded intelligence is becoming table stakes. As users put more business weight on LLM output, from client reports to employee onboarding, the question shifts from “Can it talk?” to “Can we trust what it says?”

Final thoughts

Today’s AI movement across the enterprise plays out like a quiet revolution. Agents that live inside your ERP. Copilots shaped for your job. Metrics tied to EBIT instead of engagement. While the splashy demos still get the headlines, the real action is under the hood, where AI is beginning to impact workflows, workforce structures, and software itself.

For founders, this is a blueprint to build on, embedded, role based, and highly verticalized AI is winning. For operators, the bet is that tightly scoped use cases tied to business units will outperform glossy platforms with little grounding. And for investors, it’s time to start asking the hard question, where’s the transformed workflow?

Until tomorrow,

- Aura